Dynamic Characteristics Analysis And Experimental Identification Method Of Complex Mechanical Systems

In recent years, along with advance of science and technology and development of production, high speed, high efficiency and high power become an important development direction of electromechanical product. This has led to the structure of heavy mechanical systems is becoming more complex, which includes various types of nonlinear links. If we still use the linear method of analysis, design, monitoring and fault diagnosis, the characteristics of nonlinear will be ignored that are closely related to the system characteristics. This will result in unacceptable mistake, and cause failure of analysis, design, operating monitoring and fault diagnosis. Because of complexity and diversity of the issues about dynamic characteristics, there can be no the general solution. The specific problem is solved by studying respective scheme and technology.This paper studies dynamic characteristics and fault diagnosis of complex mechanical systems (including the complex electromechanical and electro-hydraulic systems). Based on measured vibration data, statistical analysis, nonlinear characterization, artificial intelligence identification theory and method are studied. Author proposes preconditioning of vibration signal about complex mechanical systems and accurate digital integration method, and applies it to phase space description of nonlinear vibration characteristics; propose the improving accuracy method of neural network identification about the nonlinear system, and nonlinear feature extraction and fault diagnosis method based on artificial intelligence model. On this basis, propose the synthesis technology route of dynamic characteristics analysis, modeling and fault diagnosis about complex engineering structures and systems (such as large vibrating screen and hot-rolling mill). Solved the technical problems of dynamic analysis and diagnosis for the vibrating screen and hot mill, and gained real economic benefit.Because of its complexity of composition and process, statistical description about specific nonlinear characteristics of complex mechanical system including nonlinear links is obtained through measured data. Considering the signal distortion on account of noise and the further distortion and drift of integral, author propose the digital integration method about broadband filtering based on wavelet transform , FFT transform etc. First and second integral about the measured signal of ICP accelerometer are computed. More accurate velocity and displacement signal are obtained. On this basis, author proposes rational method, which realizes phase-space description and feature extraction of nonlinear time domain vibration signal.Founding accurate analytical model of complex mechanical systems is very difficult. Author studies the method that establish more accurate model of these systems through neural network identification. Based on the model goodness of fit, model generalization ability and residual test, a NNARX identification model is chosen. Affirm the activation function, hidden nodes and historical data and delay of neural network model. Finally obtain neural network model suited to the system on the higher accuracy. This provides conditions for the follow-up analysis of the nonlinear characteristics.Operating status and rules of the complex electromechanical and electro-hydraulic systems are extremely complicated. Under many cases the analysis based on dynamic equations (linear or nonlinear) can not accurately reflect its movement. Author combines the data acquisition, phase space reconstruction with identification of nonlinear systems. Phase space reconstruction and feature extraction are accomplished by using aforementioned data preprocessing and neural network identification method; Propose three-dimensional spectral analysis method based on NN identification model, that is, take some specific virtual exciting signals as input of the more accurate nonlinear system neural network model, and through simulation output obtain three-dimensional spectral that reflect the characteristics of the system. This method analyses dynamic characteristics under high credibility, and the results can be used in fault diagnosis of complex systems directly or indirectly.Analyze the operating characteristics of large vibrating screen and diagnose crack using aforementioned theoretical and experimental research results. Based on operating characteristics analysis of measured data, Author finds that the vibration of vertical on sides is chaos, and chaos degree in different locations is different. Find the path of the system leading to chaos though the simulation output signal of neural network identification model. Crack of scale models is detected by three-dimensional spectral.Finishing mill unit is a complex electro-hydraulic system. When Electro-hydraulic servo hydraulic pressure system stability lower because of fault or improper design, frame together with other mechanical structure generate intense vibration. Intense vibration impacts on product quality and safe operation of equipment. Author proposes integrated technology method including measurement, and vibration spectrum analysis, phase space analysis, system identification and stability analysis of pressure control system, and solves the problem about strong vibration of hot rolling mill.